In order to overcome the high iteration load and error diffusion existing in maximum entropy method (MEM) and to improve the prediction accuracy of ESPＲIT algorithm subjected to autocorrelation function estima-tion，a flat fast-fading channel prediction algorithm on the basis of extended Prony algorithm is proposed． By means of extended Prony algorithm，complex roots can be obtained via least square fitting and higher algebraic equations without estimating autocorrelation function，batch computation becomes possible，and thus both the prediction accu-racy and the computation efficiency improve． Theoretical and simulated results show that，in comparison with MEM and ESPRIT algorithms，extended Prony algorithm provides higher prediction accuracy，3 ～5 dB advantages at the same prediction accuracy，more steady performance and higher operation efficiency．